An improved UKF algorithm for error reduction in the robot s interior self-localization was presented by combining minimum skewness sampling and fading memory filtering methods. The algorithm adopted minimum sknewness sampling strategies for improving real time performance, decreasing the number of sampling points from 2n + 1 to n +2. Adopting the fading memory square root filter to correct the weight values of measurement noises, so as to weaken filter divergence and improve robustness of the system. Secondly, the wireless local area network (WLAN ) localization system was constructed and the improved UKF algorithm helped to correct the received signal strength indicator ( RSSI) values from wireless router. Lastly, using a trilateral calculation method for the robot coordinate , results showed the localization system reduced average localization deviation by 49% with an average value of 0.505m, which satisfies the accuracy position of the indoor-robot requirements.%为解决移动机器人室内定位误差较大的问题,提出一种将最小偏度采样策略和衰减记忆滤波相结合的改进UKF(unscented Kalman filter)算法.该算法采用最小偏度采样策略,采样点个数由2n+l减少到n+2,提高了定位实时性;采用衰减记忆平方根滤波修正量测噪声的权值,避免滤波发散,提高了系统鲁棒性.构建无线局域网定位系统,使用改进的UKF算法对获得的无线信号(RSSI值)进行滤波.采用三边定位法进行定位计算.实验结果表明,系统平均定位误差降低49％,达到0.505 m,可较好地实现机器人的精确定位,满足移动机器人的室内定位要求.
展开▼
